Implicaciones para el futuro

INTRODUCTION

Bi-parental care is the most common form of parental care in birds (Cockburn 2006). The extent of bi- parental care at different stages of breeding e.g. nest building, incubation and the provisioning of nestlings, varies between species, but both parents often contribute to nestling provisioning in altrical species. The combined provisioning efforts of both parents has been found to be related to brood mass and the fledging success of nesting attempts in several studies (Whittingham et al. 1994, Sánchez

et al. 2018). However, successful reproduction has also been widely found to be possible for single

female parents via male removal experiments, as well as in polygynous study systems, where some females receive no ‘help’ from male partners (Bart and Tornes 1989, Duckworth 1992, Gowaty 1996, Sejberg et al. 2000). This raises the question of the value and function of paternal care. It has been theorised that paternal care may only be necessary for successful reproduction in relatively poor conditions and that the costs to males of additionally providing care in good conditions may be so low that there has not been strong selection for condition-dependent care (Bart and Tornes 1989, Duckworth 1992). In addition, paternal care may increase the likelihood of having future broods, either by maintaining the females’ condition which may improve the probability of her laying an additional clutch or by increasing the females’ willingness to remain with the male for an additional clutch, rather than mate-switching (Bart and Tornes 1989).

There is a growing amount of research on multi-brooded species, showing that multi-brooding individuals can achieve greater annual and lifetime reproductive success (e.g. Chapter 3, Weggler 2006, Townsend et al. 2013, Hoffmann et al. 2014, Cornell and Williams 2016). The vast majority of studies of the determinants of multi-brooding have reported that the timing of the first brood is the most important factor (e.g. Chapter 3, Weggler 2006, Townsend et al. 2013, Hoffmann et al. 2014). Several studies have additionally reported that female traits influence the probability of multi- brooding (Geupel and DeSante 1990, Weggler 2006, Bulluck et al. 2013, Hoffmann et al. 2014), while very few studies have even considered the potential influence of male traits. A recent study of Japanese tits, Parus minor, found that the contribution of males to overall first brood nest provisioning rates was positively associated with the probability of double brooding (Nomi et al. 2018). However, previous work on dark eyed juncos, Junco hyemalis, found male provisioning care (presence versus absence of male care) had no effect on the probability of females double brooding (Wolf et al. 1991).

83 In this study, the extent of male reed warbler, Acrocephalus scirpaceus, contributions to the provisioning of nestlings on the probability of double brooding was assessed. The influence of paternal investment on the current breeding attempt was not assessed, as within breeding attempts, reduced provisioning effort by one parent can be compensated for by the other (e.g. Whittingham et al. 1994, Sandell et al. 1996, Sejberg et al. 2000). Instead, the prediction was that increased male parental investment during first broods, would reduce the workload of females, resulting in improved female condition and consequently an increased propensity to attempt a second brood. The difference in parental investment by males and females across the breeding season was also assessed. Other studies report that male and female parental investment during nestling provisioning may differ as a result of different foraging techniques, with for example, the sexes bringing different sized prey deliveries to nestlings (Sejberg et al. 2000, Krupa 2004, Krystofkova et al. 2006, Falconer et al. 2008, García-Navas et al. 2013), but it is not clear whether this is consistent seasonally. Hypothetically, if a function of male care is ultimately to increase the probability of the female initiating another clutch, male provisioning effort may decline during the breeding season relative to female effort, as the probability of double-brooding is known to decline seasonally (Chapter 3).

METHODS

Reed warblers are multi-brooded (Chapter 3), predominantly monogamous (Halupka et al. 2014) and exhibit bi-parental care throughout the nesting cycle. Both sexes contribute to incubation and nestling provisioning duties (Brown and Davies 1949, Klimczuk et al. 2015). The species is a generalist insectivore that breeds in wetland habitats, which are generally considered to provide abundant insect food resources throughout the breeding season, as there are sequential peaks in the population sizes of different insect taxa (Halupka et al. 2008, Both et al. 2009).

Data collection

All nests were systematically found and monitored in a population of approximately 150 pairs of reed warblers at Watermill Broad Nature Reserve, Norfolk, UK, throughout the 2014-15 seasons. Nests were checked regularly to determine first egg dates (FEDs), brood size and ultimate success (taken as the fledging of one or more nestling) or failure (Chapter 2). Adults were colour-ringed and also sexed, whenever possible (Chapter 2). Parent identities at nests were ascertained by setting up a video camera, on a tripod, approximately 2-3 metres away from nests. Reeds obscuring the view of the nest from cameras were tied back with wire and videos lasted approximately one hour. Nest videos were undertaken only at the nestling stage in 2014, but from the incubation stage in 2015 and subsequent

84 video attempts were made at active nests until both parents were identified. All videos were carried out in fair weather.

Provisioning data were collected from nests of pairs where both parents were known to be colour- ringed (although this sample was supplemented for one analysis, see data analysis) and which were video recorded when nestlings were day 4-6 (day 1 = hatch day). Most nestling stage videos were carried out at this time as nestlings were ringed at this age and video recording was often undertaken immediately prior to nestling ringing. Some individuals showed avoidance behaviour after the camera was set up, before resuming provisioning. A preliminary analysis (ANOVA) of the rate of provisioning visits in 10 minute time blocks showed that provisioning rates of both sexes were lowest in the first 10 minutes, significantly so for males only (p = 0.018, n = 10) and there were no significant differences between any other 10 minute time blocks (50 minutes of 10 nests were assessed). Data were therefore collected from 10 minutes after the camera was set up, for both sexes. Data were collected for up to 50 minutes thereafter (or until the end of the video if it was shorter than 60 minutes in total) and an hourly provisioning rate was then calculated. Provisioning rates of both parents were calculated by recording the time and parent identity, identified from rings, of each food delivery. Identification of adults based on differences in facial plumage markings was additionally necessary for a very small minority of visits e.g. in cases where a reed stem completely obscured the view of the colour-ring. Ultimately, all provisioning visits were successfully assigned to individuals. The bolus size of food deliveries was also classified as small or large (taken as being less than or greater than the size of the parents’ bill) and prey was identified in rare cases where this was possible (see Appendix 4). The extent of brooding behaviour of both sexes was also recorded.

Weather data from the nearest Met Office climate station at Santon Downham, 8.5 km southeast of the study site was used in analyses (as in Chapter 3). Food availability (invertebrate abundance) was measured via the weekly collection of invertebrate samples from seven water traps placed within reed beds during both study years, following the methods used by Bibby and Thomas (1985, see Chapter 2).

Data analysis

Statistical analyses were performed in R version 3.5.0 (R-Core-Team 2018), using ‘dplyr’ for data manipulations and ‘ggplot2’ for the production of figures (Wickham 2009, Wickham et al. 2018). In all analyses, numeric variables were centred and scaled and collinearity between explanatory variables (fixed effects) was assessed via pair-wise correlations and variance inflation factors (VIFs; Zuur et al. 2009). Data from nests in which one or both parents still exhibited camera shyness/nest avoidance during the data collection period (i.e. after the first 10 minutes of videos) were excluded from all

85 analyses. Camera shyness/nest avoidance behaviour was considered present for an individual fulfilling one or more of the following criteria: an individual with zero provisioning visits during the video; an individual which persistently alarmed for long periods of the video; or an individual that was observed carrying food but not delivering it to the nest (either as a result of not approaching the nest or approaching and retreating). Data from 16 nests were excluded on this basis (camera shyness being classified for the male only, female only or both parents for nine, five and two nests respectively). To assess the potential for male effort to influence the likelihood of double brooding, the probability of attempting a second brood after fledging a first brood (1 = yes, 0 = no) was modelled with a general linear model, with a binomial error structure and a logit link function. This analysis was therefore carried out on successful first brood nests only. The number of first brood nests with both parents colour-ringed available for this analysis was insufficient, so the sample was supplemented with nests from 2014, for which only one parent was colour-ringed (and the sex was known) and the other parents’ status was also known i.e. whether it was not ringed or had a metal ring only. This increased the sample size from 25 to 38 nests, however, one of these was excluded as it was part of a case of divorce, whereby the female did re-nest after the first brood, but with another male. The sample used in the analysis was therefore 37 first broods. Identification of nests as first brood attempts as well as whether or not pairs were ultimately double-brooded or not, was determined from complete breeding histories of all pairs, completed using a relative proximity method of nest ownership assignment (Chapter 2). Individual identity could not be fitted as a random effect as a result of the presence of individuals which were not colour-ringed in the sample. However, pseudo-replication, as a result of individuals being present in the sample more than once, was considered to be minimal, as only four males were known to be present in both study years in the sample used in this analysis (n = 37 colour- ringed male observations), and of these, three were double-brooded in one but not the other year. It therefore seems unlikely that individual effects would have a large influence on the probability of double brooding in this sample. The main fixed effect was the proportion of total nest provisioning contributed by the male. While the absolute rate of provisioning would be influenced by factors, which were not consistent between video recordings, such conditions would be experienced equally by both parents, so the proportion of visits by the male was taken as representative of his level of investment. Other fixed effects comprised explanatory variables already known to influence the probability of double brooding in the study population: first egg date, food availability, temperature, rainfall and brood size (Chapter 3). Temperature (⁰C) and rainfall (mm) were taken as the mean and total for the five days following the fledge date of nests, as in Chapter 3. Food availability was taken as the total invertebrate mass of the seven water traps, in the weekly sample collected on the date on or following the fledge date of nests (as in analyses in Chapter 3). First egg date and temperature were collinear (r

86 = 0.7), however, both were retained in the model, in order to be able to account for all explanatory variables previously identified as important predictors of double brooding. All other pair-wise correlations between fixed effects were relatively low (r values ≤ 0.5, VIFs < 2.7) and the model was validated with a binned plot of the average residuals and fitted values.

The effect of nest timing on the provisioning rate of individuals (the number of deliveries per hour) was modelled using a linear mixed effects model. This analysis was undertaken only on nests for which both individuals were colour-ringed (n = 80 individuals from 40 nests). Individual identity, nested within pair identity was included as a random effect, to control for the occurrence of multiple nests of the same pair within a season. Fixed effects included date and sex as well as year (there being insufficient levels to fit year as a random effect; Bolker et al. 2009), brood size (on the day of video recording), time of day and a measure of food availability as well as the interaction between sex and date, in order to assess if seasonal trends in provisioning rates differed between females and males. Time of day was included as a categorical variable, with early, mid or late timing of videos determined from whether the nearest hour to the start of data collection period (i.e. 10 minutes after the start of the video) was earlier than 11 am, between 11 am and 2 pm or after this period. Food availability was included in order to control for the influence of resources on provisioning rates and was taken as the total invertebrate mass of all water traps on site, collected on the sample date equal to, or the nearest following, the video recording date. All pair-wise correlations were relatively low; r values ≤ 0.5 and VIFs < 1.7. Plotting the model residuals against the fitted values, identified one outlier, so the model was run both with and without the outlier (without the data from either individual from the nest). There were no fundamental differences between the predictions from either model and the model including the outlier is reported in the results.

The effect of nest timing on the size of food deliveries of individuals, was modelled using a generalised linear mixed effects model. The proportion of deliveries which were classified as large bolus sizes, out of all deliveries to the nest by each individual, was modelled with a binomial error structure and a logit link function. The same sample and model structure was used as for the provisioning rate analysis. The model was validated with a binned plot of average residuals versus fitted values.

For all analyses, models were run using ‘lme4’ (Bates et al. 2018) and ‘MuMIn’ was used to fit and rank all candidate models by AICc values (sample size adjusted Akaike Information Criterion; Bartoń 2018). The best-fitting models, those within 2 AICc units of the top model, were averaged for figures of modelled effects.

87 RESULTS

Provisioning rates (feeds per hour) were similar between females (mean = 9.0, sd = 4.5) and males (mean = 10.4, sd = 4.8, n = 40 colour-ringed pairs) and showed no clear seasonal trend in either sex according to the raw data (Figure 5.1). Of the 40 colour-ringed pairs, 24 females brooded the nestlings during videos (range = 1-39 minutes), while seven of the males undertook any brooding (range = 4-32 minutes per hour). The proportion of first brood provisioning undertaken by males varied from 22-93 % (mean = 53%).

Of the 37 first broods included in the model of double brooding probability, 20 were of pairs which were ultimately double-brooded. The model identified similar predictor effects as documented previously in Chapter 3, although the previously documented effects of food availability and temperature were both absent in the best-fitting models (Table 5.1). The effect of temperature was likely absent as a result of being collinear with FED, the most important predictor of double brooding, while the effect of food availability could have been absent as a result of the study years used in the analysis representing the seasons with most abundant food resources (compared to additional seasons used in analyses in Chapter 3). The probability of double brooding was predicted to increase with an increasing proportion of the total nest provisioning contributed by the male (Table 5.1), however, this effect was small and the prediction confidence intervals were very wide, so there is insufficient evidence for a real effect (Figure 5.2).

The best-fitting models of provisioning rate, contained sex, date and the interaction between these terms (Table 5.2): all three of these effects were weak (Figure 5.3). Seasonal variation was minimal and although there was a suggestion that males may provision at greater rates, the confidence intervals for the predicted rate for females and males overlapped. Sex was an important predictor of the proportion of large bolus deliveries, males providing a greater proportion (Table 5.3, Figure 5.4). As with provisioning rates, the proportion of large bolus deliveries did not change seasonally. Date was present in one of the best-fitting models but had a very small effect size, while the sex-date interaction was absent. Food availability had a small effect on the proportion of large deliveries but only a negligible effect on provisioning rates (Table 5.2, Table 5.3).

DISCUSSION

The hypothesis that male contribution to nestling provisioning would influence the probability of double brooding, was not supported by this study. Very few preceding studies have assessed the role of paternal care in influencing the likelihood of multi-brooding. In agreement with the current study,

88 Wolf et al. (1991) reported, in dark-eyed juncos, that male contribution to provisioning had no influence on double brooding, as females whose partners were experimentally removed, were just as likely to initiate additional clutches as control females. However, Nomi et al. (2018) recently reported that in Japanese tits, male contribution to provisioning did have a positive influence on double brooding. The differences between the results of these studies could be related to the fact that Japanese tits have larger brood sizes, which may mean there is greater scope for costs of reproduction to females whose partners contribute less to care (Nomi et al. 2018). In reed warblers, male contribution to provisioning may not be necessary for successful reproduction and therefore variance in male effort may not impact upon female condition to the extent that influences her likelihood of laying another clutch. Indeed, it has been found in reed warblers that successful reproduction, the fledging of nestlings, is achievable for single females, though male care may be necessary at the end of the season when conditions may be poorer (Duckworth 1992). However, many studies have found a cost of reproduction (references within Chapter 4) and in fact, even in Wolf et al.’s (1991) study, despite single females being just as likely to initiate second broods, they did suffer a cost in body condition. Females paired with males which contribute less may therefore still suffer a cost, alternative to a reduced likelihood of multi-brooding.

The current study likely had a relatively low ability to detect a small effect of male contribution, as a result of controlling for several other factors already known to influence double brooding in the study population. Nomi et al.’s (2018) model of double brooding was comparatively simple, however, the effect of male contribution which they found in Japanese tits is quite remarkable, as its effect was considerably more important than the timing of hatching. This is interesting, as in the vast majority of study systems, for which multi-brooding determinants have been investigated, the timing of the first brood has been found to be the most important factor (e.g. Geupel and DeSante 1990, Ogden and Stutchbury 1996, Verboven and Verhulst 1996, Brinkhof et al. 2002, Weggler 2006, Bulluck et al. 2013, O’Brien and Dawson 2013, Townsend et al. 2013, Carro et al. 2014, Hoffmann et al. 2014, Zając et al. 2015, Béziers and Roulin 2016, Jackson and Cresswell 2017).

It is necessary to acknowledge that a possible limiting aspect of the current study to detect an effect of paternal care, could have been the age of nestlings at the time of video recordings. Previous studies have found that the roles of parents can change during the course of the nestling stage (García-Navas

et al. 2012) and therefore it may be that male contribution at later stages may be more important for